3 providers are running promotional pricing right now. Two of them don't say when it ends. · Verified September 15, 2026
Evidence reviewed September 15, 2026 · 34 sources
Almost every article warning that GLP-1 drugs “eat your muscle” is reading a number off a body scan that was never a muscle measurement. Getting that one distinction right changes the answer, and changes who actually needs to worry.
A body-composition scan sorts your tissue into three buckets: fat, bone mineral, and everything else. That third bucket is called lean mass, and it contains your skeletal muscle — along with your liver, kidneys, gut, blood, and the water bound up in all of it. Glycogen alone holds roughly three grams of water per gram of glycogen.
So when a trial reports “lean mass fell 9.7%”, it has not measured your muscle. It has measured muscle plus organs plus fluid, and it cannot tell you how the loss divided between them. In one 2026 study of obese mice on these drugs, the drop in liver mass was larger than the change in muscle.
The four-society nutrition advisory that most clinicians follow puts actual muscle at about 20% of weight lost. The “40%” figure in circulation is a lean-mass number at the top of its range, being reported as though it were muscle.
Randomised trial Semaglutide 2.4 mg, 68 weeks (STEP 1 scan substudy, n=140). Body weight −15.0%, fat mass −19.3%, lean mass −9.7%. That works out to roughly 38% of the tissue lost being lean mass. One caveat that matters and is almost never mentioned: this substudy exists only as a 2021 conference abstract. It has never been published as a peer-reviewed paper, and it is the single most-quoted result in the entire debate.
Randomised trial Tirzepatide, 72 weeks (SURMOUNT-1 scan substudy, n=160). Weight −21.3%, fat −33.9%, lean mass −10.9% — about 25% of weight lost, as stated by the authors.
Randomised trial Pooled across 22 trials (network meta-analysis, 2,258 people). Lean mass −0.86 kg (95% CI −1.30 to −0.42), about 25% of total weight lost. Lean mass as a proportion of your body did not change.
Across the literature the honest range is 10% to 60%, clustering at 25–40%. Anything quoting you a single number is hiding the spread — which varies by drug, dose, starting body composition, sex, age, and which machine did the measuring.
This is the question that actually matters, and it is the one almost nobody asks. Losing weight always costs lean tissue. The long-standing rule of thumb is that about a quarter of any weight lost is fat-free mass.
Randomised trial The best direct test we have found no disproportionate loss. SURPASS-3 MRI measured actual thigh muscle volume in 246 people on tirzepatide for a year, then compared it against what 2,942 UK Biobank participants' own data predicted for that much weight loss. Muscle volume fell 0.64 L — and the difference from expectation was −0.04 L (95% CI −0.11 to 0.03, p=0.22). Statistically indistinguishable from normal. Muscle quality improved: fat infiltration fell more than expected (p<0.0001). One signal against — at the 15 mg dose, muscle volume did fall more than predicted (p=0.0016). This trial was funded by Eli Lilly, who make the drug.
For comparison: bariatric surgery costs roughly 8 kg of lean body mass at twelve months, over half of it in the first three. Ordinary dieting runs 16–23%. Normal ageing takes 3–5% of muscle per decade after 30, and strength faster than that.
Here is the finding that the coverage gets backwards. The alarm is broadcast at everyone. The measured functional harm shows up in one group.
Randomised trial In adults over 60, function measurably declined. In the placebo arm of a phase 2b trial — semaglutide alone, 16 weeks, all participants aged 60+ — 42.6% had a drop of 10% or more in stair-climbing power. A median 32% of their weight lost was lean mass. For scale, stair-climb power normally declines about 1.4% a year with age, so a 10% drop is roughly seven years' worth. Sponsor-reported topline result, not yet peer-reviewed.
Observational And it persisted at two years. In 220 semaglutide patients aged 65+ matched against 212 controls, grip strength was 1.6–2.5 kg lower and gait speed measurably slower at 24 months (all p<0.05). Real-world records, so it cannot prove cause.
Randomised trial In younger adults, strength held up. Trials of semaglutide and liraglutide have shown preserved grip strength despite falling lean mass. In one 12-month cohort of 106 patients, grip strength increased and the share meeting criteria for sarcopenic obesity fell from 49% to 33%.
The reviewer who put it best: lean tissue loss “is not a reliable predictor of muscle strength change.” Losing lean mass on a scan and losing what you can physically do are not the same event, and in people under about 60 the second one largely has not been observed.
Randomised trial Exercise alongside the drug. The strongest card, and weaker than you would expect. A 195-person, one-year trial found the combination of exercise and liraglutide cut body-fat percentage by 3.9 points — roughly double either alone — and only the combination improved blood sugar control and fitness. Adding exercise also abolished the bone-density loss seen with the drug alone. Two caveats the headlines drop: that programme was mostly aerobic, not resistance training, and liraglutide is the weakest drug in this class. No completed randomised trial has tested resistance training during semaglutide or tirzepatide. The first one designed to measure it properly is running now and reports later.
Observational Protein. Real guidance, extrapolated evidence. The joint advisory of four US professional bodies proposes 1.2–1.6 g per kg per day during active weight loss, with strength training at least three times a week plus 150 minutes of aerobic exercise. European guidance for older adults says at least 1.0–1.2 g/kg and at least 25 g per meal. The catch that never makes the headline: the 2026 international consensus behind these numbers states plainly that its recommendations are “primarily derived from indirect evidence” — borrowed from bariatric surgery nutrition — and that direct evidence is “urgently required”. No trial has tested whether hitting a protein target during GLP-1 treatment preserves anything.
This section is here because it is the reason to believe the rest of the page.
Unsupported Creatine protects your muscle on a GLP-1. There are zero completed trials in this population. The first — a 40-person pilot — began in May 2026 and does not report until 2027. Creatine is genuinely well evidenced in resistance-training populations; carrying that across to this one is an assumption being sold as a finding.
Unsupported HMB, BCAAs, collagen and “muscle-preserving” blends. We found no trials of any of them in people taking these drugs. The four-society advisory's supplement guidance runs to vitamin D, calcium, B12, a multivitamin, and magnesium for constipation. Creatine is not on it.
If you want to know what is happening to your own body rather than to a trial average, there are three ways to look, and they are good at different things. Nobody explains the difference, so here it is.
Reference standard A DXA scan is the method the professional statements name. It is the one that produces the numbers in every trial on this page. Roughly $50–150 at a scanning centre, and the only way to get a figure precise enough to compare against a study. Availability is the catch — the scanning networks cover a few dozen metros between them. If there is one near you, a scan at the start and another at six months will tell you more than anything else here.
Useful, with limits A bioimpedance scale is not a DXA and is not trying to be. It passes a small current through you and estimates composition from the resistance, which makes it sensitive to things that have nothing to do with muscle — how hydrated you are, what you ate, the time of day. So a single reading is not a measurement of your muscle, and a 2% change between two mornings is inside the noise. What it is genuinely good for is direction over months, weighed under the same conditions each time — same time of day, before eating, similar hydration. A trend line from a $50 scale is real information. It just is not a number you can put next to a trial result.
Useful, with limits Grip strength costs about $30 and may tell you more than either. This is the finding buried in the research: strength and lean mass came apart in the trials. Adults who lost lean mass on a scan often kept their strength, and the one trial that successfully protected lean mass produced no strength improvement at all. A hand dynamometer measures the thing that actually matters to your life, and it is the cheapest instrument on this page.
Randomised trial And the assumption underneath all of it is in doubt. A 507-person trial of a muscle-preserving antibody alongside semaglutide successfully protected lean mass — participants gained 2.5% while the semaglutide-only group lost 7.4%. It produced no significant improvement in grip strength or in how patients said they felt. Preserving tissue on a scan and preserving function came apart in the first trial large enough to look at both.
Four muscle-preserving agents are in trials alongside these drugs — an anti-activin-receptor antibody from Eli Lilly, an oral drug from Veru, an anti-myostatin antibody from Scholar Rock, and a Regeneron combination that preserved about 81% of the lean mass otherwise lost.
Every one is phase 2. None is approved. None has been tested against an endpoint that is about a person's life — not disability, not fractures, not survival. And most of the numbers in circulation come from investor press releases rather than published papers.
If you are under 60: you are losing lean tissue at roughly the rate any weight loss costs, the best direct measurement found nothing unusual about it, and trials have generally found your strength holds. Train and eat protein because they are good for you, not because you have been told your muscles are dissolving.
If you are over 60: this is the group where the functional decline was actually measured, and it was substantial. Resistance training and protein matter more here, and this is a conversation worth having with the person who prescribed it.
Either way: nobody sells a supplement with evidence behind it for this. The two things every professional body recommends are lifting and eating enough protein, and even those rest on borrowed evidence.
Reviewed September 15, 2026. Every figure above was read from the published trial, meta-analysis or professional guidance cited, and each is graded for what kind of evidence it is. Where a number comes from a conference abstract, a sponsor press release, or a trial funded by the drug's manufacturer, we say so on the claim rather than in a footnote. Sources are listed in full below.
We keep a table of every provider we track, ranked by what twelve months actually costs rather than the price they advertise — including the ones that will not tell you what month two costs.
That spread moves. 3 of these providers are running promotional pricing right now, and two of them will not say when it ends. We check every week. Want us to tell you when yours changes?
Your email and nothing else. Not your weight, your dose, or which page you were reading. We deliberately do not store that. No trackers, never sold.